To allow for cases where people want to bind on all listening sockets, we have to
allow for the ...

When large numbers of mounts are defined, it was taking a long time to do the recheck
routine (w...

xml parsing can take a little time for large configs so allow things to continue
while it is bei...

these others don't need to be reported immediately

certain stats like global listener counts do not need to be immediately reported on
the log (deb...

normally it is move/delete file and then HUP (logs reopen), but if someone created
a file (typic...

If large number of listeners are on a stream and it drops then a large set of log messages occur...

Incoming clients processed by the listener thread pushes these clients to the normal worker
pool...

This reverts commit ee2258f13892bc2cf6567e163998fd17411aa40c.

type cleanups, minor range checking, etc

gettimeofday, while portable, can be affected by things like ntp in a significant
way. The repla...

could do with some reordering of clients to really be useful but for now have quite a
large cap ...

having a very short limit causes files to re-open. This just enforces a sane minimal amount

Some prefer to move/delete log files then signal but don't require the archiving aspect. The
log...

probaby minor effect but does not hurt having it.

To keep in sync with xiph master, add these to allow passing certain listener headers
to the aut...

reopening of log can expose a stale pointer deref, leading to crash or bad memory lookup

Fix a typo in an error message

lets not be too wordy in the POST args. the stats routine returns a string
that needs to be free...

Report number of users on the listener_add action

the auto log cycle based on file size was broken on async update. The counter for
buffer sizing ...

When authenticating a listener, the action 'listener_add' will now also send
the number of activ...

As we allow for updates to SSL certs over reload, there is a chance that the reload will fail
to...

Call the reading of SSL key files like we do for the socket opening. When a re-read is done
and...

just for consistency and a #ifdef change for the build

The purge routine was acting too aggressively. Depending on the exact processor
setup, the abili...

* use server cipher preference (SSL_OP_CIPHER_SERVER_PREFERENCE)
* enable all bug workarounds (S...

The purge routine uses the keep entries setting for the min value but the pop routine within
it ...

the main logs default to a number of lines for display or debug purposes but the per-mount
acces...

The ugly handling for removing entries from the cached tree is now gone. It only affected
the ba...

SSL autodetect was tripping over a case not handled, if SSL certifcates were not defined

the examples I used had this in but they were subsequently altered to automatically determine
th...

Log messages from any root startup were not logged as the log file should be as the changed owne...

It seems that with some proxies (eg aws), the read channel can be shut down while the write
chan...

the ssl-certificate option can still be used like before but some prefer
keeping the private key...

The <ssl> tag is not needed now, it is still there to enforce SSL only but it
is no longer requi...

the windows build had an issue previously with ipv6 so was disable.

This allows for the checkin...

I'll lessen the default max size as well

actually traverse the logs and quickly return if not in use.

This extends on the internal log entriy handling to allow for a separate thread to
run through t...

It has been reported that Pure radio boxes have reboot issues if the pragma header is present, w...

openssl uses some global structures which could suffer from race conditions.

no stability issue, just looks weird in the access log

The negative test needs to allow for a larger range if some large frame needs to
be relocated. I...

very small intervals from relays were not handled properly, parsing metadata
itself worked but t...

some rearragement of the files, updates to scripts themselves. I'll keep the autotools
output fo...

For relays that are created in a slave, the retry period has been traditionally set
by the maste...

This is to complement the run-on setting for relays, as master relays are dynamically
created in...

odd one to trigger this. range request using auth which returns intro content being placed on an...

It's dependent on the block reading. Normal cases will be fine but you can get the password
and ...

the intro offset is either 0 (initially), set later by the range check, or reset by the keep-ali...

CORK may already be enabled by default of course but we should drop the CORK
setting to ensure d...

If MY_ALLOC compiled in, then track these allocations and don't force the stack size on windows
...

The stats may eventually only apply to if there is a change each second but for now on large
cha...